Fault Feature Separation Of Gearbox Based On Spectral Kurtosis

Gearbox has been widely used in various fields. It is applied as the common component to transmit power and shift speed in mechanical equipments. The working condition of a gearbox can be extremely bad, which make parts of the gearbox fail eas-ily. As a result, it is significant to monitor and diagnosis gearbox in order to evaluate the running condition of a machinery. The signal processing method based on SK had been developed recently as a new faults diagnosis way. Due to its novel ability of separating the faulty signals from the noise background, it has been widely studied in fault diagnosis area. However there are still some difficulties that need to be solved before it can be put into applications. Firstly, the faulty signals in the real world are usually merged by strong noise. Secondly, although the envelop analysis method based on SK can obtain its parameters adaptively but it needs an intensive computation. Moreover, how to apply this method to the run up or run down processes for the faults diagnosis of gearboxes etc.In order to solve these problems, a pre-whiting process was did for the original signals using autoregressive (AR) modals to enhance impact signals of gear fault. Then an improved envelope analysis approach based on spectral kurtosis and complex shifted Morlet wavelet in the diagnostics of local gear faults is proposed in this paper, Using this improved means based on the relation of the filtering and the results of the spectral kurtosis to filters with an overlap filtering frequency band at different filter bands, not only both the optimal center frequency and bandwidth of the band pass filter in envelope analysis can be obtained adaptively combined with spectral kurtosis, but also the number of the required band-pass filters are reduced, so that it significantly improves the calculative efficiency and lowers the correlative cost of the computing. Simulation and tests results of gear fault verified the feasibility of the present scheme positively.Meanwhile, through the research on the gearbox vibration signals at the variable speed conditions of rotating machinery, a method of envelope order tracking analysis of rolling elements bearing faults based on the spectral kurtosis is proposed in this pa-produced by rolling elements bearing faults will excite the resonances of its or its surrounding structure. The method proposed in this paper can use the spectral kurtosis algorithm to obtain both the optimal center fre-quency and bandwidth of the band-pass filter in envelope analysis adaptively. Then the envelope signals that contain the vibration features of the initial stage of the rolling elements bearing faults can be gained. Consequently, the envelope signals are re-sampled by even-angle sampling scheme, which ensure the non-stationary signals in time-domain are turned into a quasi-stationary signal in angle-domain. As a result, the frequency smear is eliminated in order spectrum and the fault diagnosis of rolling ele-ments bearing on the variable speed conditions of rotating machinery is achieved. Simulations and tests verified the feasibility of the proposed method.